Re: Regression in serial console on ia64 after 2.6.22

[Yinghai Lu <Yinghai.Lu@Sun.COM>]

Horms wrote:
> On Tue, Jul 24, 2007 at 04:57:32PM -0700, Yinghai Lu wrote:
>> IA64
>>
>> Subject         : Regression in serial console on ia64 after 2.6.22
>> References      : http://marc.info/?l=linux-ia64&m=118483645914066&w=2
>> Last known good : ?
>> Submitter       : Horms <horms@verge.net.au>
>> Caused-By       : Yinghai Lu <Yinghai.Lu@Sun.COM>
>>                   commit 18a8bd949d6adb311ea816125ff65050df1f3f6e
>> Handled-By      : ?
>> Status          : unknown
>>
>>
>> please test this patch.
>>
>> YH
> 
> I just posted a similar patch (sorry, I didn't see your post until
> afterwards). I guess that we both have the same ideas :)
> 
>>
>> [PATCH] ia64: move machvec_init before parse_early_param
>>
>> So ia64_mv is initialized before early console
>>
>> Signed-off-by: Yinghai Lu <yinghai.lu@sun.com>
>>
>> diff --git a/arch/ia64/kernel/machvec.c b/arch/ia64/kernel/machvec.c
>> index 13df337..a94feaa 100644
>> --- a/arch/ia64/kernel/machvec.c
>> +++ b/arch/ia64/kernel/machvec.c
>> @@ -14,12 +14,6 @@ struct ia64_machine_vector ia64_mv;
>>  EXPORT_SYMBOL(ia64_mv);
>>  
>>  static __initdata const char *mvec_name;
>> -static __init int setup_mvec(char *s)
>> -{
>> -	mvec_name = s;
>> -	return 0;
>> -}
>> -early_param("machvec", setup_mvec);
>>  
>>  static struct ia64_machine_vector * __init
>>  lookup_machvec (const char *name)
>> @@ -42,6 +36,10 @@ machvec_init (const char *name)
>>  
>>  	if (!name)
>>  		name = mvec_name ? mvec_name : acpi_get_sysname();
> 
> I think that mvec_name can be disposed of all together as
> its only used by the line above and setup_mvec(), which in
> turn is only used by early_param. I got rid of mvec_name in
> my incantation of this patch, and it seemed to work quite well.

with cmd_line or NULL, mvec_nanme could get assigned.
next machvec_init(NULL) could reuse that

> 
>> +
>> +	if (!mvec_name)
>> +		mvec_name = name;
>> +
>>  	mv = lookup_machvec(name);
>>  	if (!mv)
>>  		panic("generic kernel failed to find machine vector for"
>> diff --git a/arch/ia64/kernel/setup.c b/arch/ia64/kernel/setup.c
>> index cf06fe7..b06d7b7 100644
>> --- a/arch/ia64/kernel/setup.c
>> +++ b/arch/ia64/kernel/setup.c
>> @@ -481,6 +481,9 @@ int __init reserve_elfcorehdr(unsigned long *start, unsigned long *end)
>>  void __init
>>  setup_arch (char **cmdline_p)
>>  {
>> +#ifdef CONFIG_IA64_GENERIC
>> +	char *mvstr;
>> +#endif
>>  	unw_init();
>>  
>>  	ia64_patch_vtop((u64) __start___vtop_patchlist, (u64) __end___vtop_patchlist);
>> @@ -491,12 +494,15 @@ setup_arch (char **cmdline_p)
>>  	efi_init();
>>  	io_port_init();
>>  
>> -	parse_early_param();
>> -
>>  #ifdef CONFIG_IA64_GENERIC
>> -	machvec_init(NULL);
>> +	mvstr = strstr(*cmd_line_p, "machvec=")
>> +	if (mvstr)
>> +		mvstr = strchr(mvstr, '=') + 1;
>> +	machvec_init(mvstr);
>>  #endif
> 
> Do you need to copy and truncate mvstr on ' ' in the case
> that machvec= isn't the last argument on the command line?
> I did this in my patch, though I didn't test to see if
> it was uneccessary.

or may need to update lookup_machvec...
from strcmp to strstr...

just want to the code to be less changes as possible.

YH